独立偏位修正法による多重撮影画像の特徴探索

摘要

Recently, large amounts of photographs are able to be obtained by large format digital cameras easily and at a low cost compared to film type cameras. As a result, the large format digital cameras in the field of aerial photograph surveying have become wide spread in the industry. Consequently, multi-view photographs that are taken from various view-points of the same area of the ground are frequently duplicated. For the improved efficiency of the multi-matching theory, we propose a new efficient feature searching method using multi-view images. By using the independent rectification images as to obtain the projection conversion to vertical direction, it will avoid the complicated 3-dimensional ray tracing problems. Therefore, it means that it is possible to realize the multi-images matching only using magnification from FOE (Focus of Expansion) of each independent rectification image as a 2-dimentional image conversion. This paper presents the geometrical specifications that make the best use of image magnification to solve the problems of ordinary multi-matching theory, such as, "decrease of matching score depends on the flight attitude", "occurrence of lacking gap depends on the occlusion" and "increment of repetitive operation for calculation costs". This independent rectification method proves to be valid for these problems, and the effectiveness is verified with the experimentation using multi-view images of actual aerial photographs.%近年，国土基盤データとしての地理空間情報に関す るニーズは多岐に渡り，地理情報システム（GIS）に留 まらず，インターネットを含む情報化社会において， インフラとしての社会的なニーズが高まっている。特 にその骨格を成す都市域における建物形状の計測技術 は重要であり，より広域な空間データを精密かつ迅速 に取得できることが望まれている。